Automatic remotely sensed image classification in a grid environment based on the maximum likelihood method

Grid technology is now being increasingly applied in the process of remotely sensed images due to its characteristics, including distribution, high performance, dynamics, extendability and full sharing of resources. Numerous pixel-based classification algorithms have been developed, many of which yield certain optimizations and mature gradually. Maximum likelihood classification (MLC) is the most widely used method. The objective of this paper is to improve the efficiency of MLC using grid technology and realize its automation with the help of a sample library which is in the form of an XML file. MLC is implemented both in the grid environment and in the stand-alone environment using a Landsat-4 TM image. Through the assessment of accuracy and a comparison between classification results obtained from ENVI 4.5 and our program, it can be concluded that the MLC algorithm is implemented well and precision is guaranteed using the sample library. Furthermore, the time cost of classification in the grid environment is greatly reduced compared with that in the stand-alone environment, which proves that the efficiency of MLC has obviously improved. (C) 2011 Elsevier Ltd. All rights reserved.